A system and method performs multilingual wakeword detection by determining a language corresponding to the wakeword. A first wakeword-detection component, which may execute using a digital-signal processor, determines that audio data includes a representation of the wakeword and determines a language corresponding to the wakeword. A second, more accurate wakeword-detection component may then process the audio data using the language to confirm that it includes the representation of the wakeword. The audio data may then be sent to a remote system for further processing.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
5. The computer-implemented method of claim 4, wherein the first wakeword-detection component comprises the classifier.
This invention relates to a computer-implemented method for improving wakeword detection in voice-activated systems. The problem addressed is the inefficiency and inaccuracy of traditional wakeword detection methods, which often suffer from false positives or missed activations due to background noise or similar-sounding words. The method involves a first wakeword-detection component that includes a classifier. This classifier is specifically designed to analyze audio input and determine whether it contains a wakeword—a predefined phrase that triggers a voice-activated system. The classifier may use machine learning techniques, such as neural networks, to distinguish between the wakeword and other sounds or speech. The first wakeword-detection component may operate in conjunction with a second wakeword-detection component, which may use a different detection technique, such as template matching or keyword spotting, to enhance accuracy. The system may also include a confidence scoring mechanism to evaluate the likelihood that detected audio segments contain the wakeword, allowing for more reliable activation of the voice-activated system. The method aims to improve the robustness and efficiency of wakeword detection in noisy environments or when dealing with similar-sounding words.
15. The system of claim 14, wherein the first wakeword-detection component comprises the classifier.
A system for wakeword detection in voice-activated devices includes a classifier that processes audio input to identify a predefined wakeword. The classifier is trained to distinguish the wakeword from other sounds, improving accuracy and reducing false activations. The system may also include a pre-processing module that filters and normalizes the audio input before classification, enhancing detection performance. The classifier may use machine learning techniques, such as neural networks, to analyze audio features and determine whether the wakeword is present. The system may further include a post-processing module that validates the detection result, reducing errors from background noise or similar-sounding words. The wakeword-detection component operates in real-time, allowing the device to respond quickly to user commands. The system may be integrated into smart speakers, virtual assistants, or other voice-controlled devices to enable hands-free operation. The classifier's design ensures low-latency processing while maintaining high accuracy, addressing challenges in noisy environments or with varying accents. The system may also include adaptive learning mechanisms to improve detection over time based on user interactions. This approach enhances user experience by minimizing false triggers and ensuring reliable activation of voice commands.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
June 28, 2021
May 28, 2024
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.